WHAMR!: Noisy and Reverberant Single-Channel Speech Separation

By Matthew Maciejewski, Gordon Wichern, Emmett McQuinn, Jonathan Le Roux. 2019

227 citations recorded by the source corpus.

While significant advances have been made with respect to the separation of overlapping speech signals, studies have been largely constrained to mixtures of clean, near anechoic speech, not representative of many real-world scenarios. Although the WHAM! dataset introduced noise to the ubiquitous wsj0-2mix dataset, it did not include reverberation, which is generally present in indoor recordings outside of recording studios. The spectral smearing caused by reverberation can result in significant performance degradation for standard deep learning-based speech separation systems, which rely on spectral structure and the sparsity of speech signals to tease apart sources. To address this, we introduce WHAMR!, an augmented version of WHAM! with synthetic reverberated sources, and provide a thorough baseline analysis of current techniques as well as novel cascaded architectures on the newly introduced conditions.

Subfields: Artificial Intelligence, Computational Mechanics, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Mirco Ravanelli
2. Yoshua Bengio
3. Zhuo Chen
4. Titouan Parcollet
5. Shinji Watanabe
6. Lei Xie
7. Jinyu Li
8. Dong Yu

In Computational Mechanics

1. Zhuo Chen
2. Szu-Wei Fu
3. Chien-Feng Liao
4. Lei Xie
5. Jun Du
6. Emmanuel Vincent
7. R. Mori
8. Yoshua Bengio

In Signal Processing

1. Mirco Ravanelli
2. Yoshua Bengio
3. Titouan Parcollet
4. Zhuo Chen
5. Samuele Cornell
6. Emmanuel Vincent
7. Shinji Watanabe
8. Lei Xie

Citations in the corpus

• Single Channel Voice Separation for Unknown Number of Speakers Under Reverberant and Noisy Settings (2020, neutral): “While [22] reported results on WHAMR! dataset [23], which contains noisy reverberant mixtures, this dataset consists of mixtures of two sources only.”
• Voice Separation with an Unknown Number of Multiple Speakers (2020, neutral): “, 2019) and WHAMR! (Maciejewski et al., 2020) benchmarks.”
• Phase-aware Single-stage Speech Denoising and Dereverberation with U-Net (2020, methodology): “For test, we used two datasets such as the synthesized testset in the DNS challenge (DNS) and WHAMR [20].”
• SpeechBrain: A General-Purpose Speech Toolkit (2021, methodology): “ConvTasNet[43] DualPath RNNs [44] SepFormer [45] WSJ-mix [46] WHAM [47] WHAMR [48] LibriMix [49] Spoken language understanding Speech to intent/slots.”
• AISHELL-4: An Open Source Dataset for Speech Enhancement, Separation, Recognition and Speaker Diarization in Conference Scenario (2021, methodology): “[15] M. Maciejewski, G. Wichern, E. McQuinn, and J. Le Roux, “WHAMR!”
• LibriMix: An Open-Source Dataset for Generalizable Speech Separation (2020, background): “[10], respectively the noisy and reverberant extensions of wsj0-2mix.”
• Wavesplit: End-to-End Speech Separation by Speaker Clustering (2020, unknown): “This information is provided by most speaker separation datasets, including Wall Street Journal variants (Hershey et al., 2016; Wichern et al., 2019; Maciejewski et al., 2019), meeting recordings (Mccowan et al., 2005) or cocktail party recordings (Barker et al., 2018).”